Besides the surface soil moisture and temperature, sub-surface profiles of these variables play very important but still poorly understood roles in controlling the surface fluxes of water and energy. Exfiltration and infiltration fronts are partially controlled by the state of the soil below the surface. In addition, the diurnal amplitudes of surface flux and state variations are affected by the conditions below the surface. There are, however, severe limitations on monitoring conditions below the surface. In-situ point measurements are inadequate for characterizing large-scale fields, and remote sensing techniques cannot provide direct observations of moisture and temperature below the surface layer.
In order to improve our understanding of the land-atmosphere exchange processes on a regional scale, dynamically consistent estimates of profile soil moisture may be derived through the assimilation of remotely-sensed passive microwaves (as observed in SGP97) into a model for moisture and temperature dynamics. Furthermore, the problem of down-scaling low-resolution remote sensing observations is addressed in the context of merging models and measurements at different scales.
The dynamical model describes the most important near-surface and profile water and energy transport processes. Assuming predominantly vertical fluxes, we divide the computational region into one-dimensional vertical columns. Moisture transport in each column (or pixel) is described with Richards' equation while energy transport is modeled with the Force-Restore approximation of the heat equation. The meteorological forcings onto different pixels are taken from Oklahoma Mesonet data. The measured L-band brightness temperatures are related to the surface soil moisture and temperature through a simple, non-coherent Radiative Transfer model.
There are structured spatial features in the model and the micrometeorological forcing that contribute to errors. Such model errors are accounted for as uncertainties in the forcing and assumed to be random fields which are correlated over time and space. The estimates of soil moisture and temperature are derived from a variational least-squares algorithm with the dynamical model imposed as a weak constraint. Through its implicit propagation of the error covariances, the algorithm is very efficient and thus able to provide optimal estimates without the simplifications that are needed in large-scale Kalman filtering applications. Posterior variance calculations allow for a thorough test of the algorithm's underlying statistical assumptions.
In our application, we simulate the down-scaling situation encountered
when satellite data will eventually be available.
Expected satellite footprints are on the order of 35km by 35km,
whereas we estimate soil moisture on the much smaller scale
of soil and landcover data.
To mimic satellite measurements, we aggregate SGP97 airborne
measurements of brightness temperatures.
SGP97 gravimetric soil moisture data provide an excellent source of
ground-truth observations.
We present preliminary results as well as an assessment of the
computational and operational feasibility of the approach.